PGI User Forum
 SearchSearch   MemberlistMemberlist     RegisterRegister   ProfileProfile    Log inLog in 

CUDA-x86.

pgcc compile error in OpenACC-CUDA interoperabily example

 
Post new topic   Reply to topic    PGI User Forum Forum Index -> Accelerator Programming
View previous topic :: View next topic  
Author Message
i_alex2004



Joined: 18 Aug 2012
Posts: 8
PostPosted: Wed Dec 19, 2012 10:16 am    Post subject: pgcc compile error in OpenACC-CUDA interoperabily example Reply with quote
Hello, I have written a code to show OpenACC-CUDA interoperability and have some problems with compilation, could you please help me in finding problem?

Code:

#include <stdio.h>
#include <cuda.h>
#include <curand_kernel.h>

static const int N=100;

void vecAdd (float restrict *a, float restrict *b, float restrict*sum)
{
   #pragma acc kernels loop present (a[N], b[N], sum[N]);
   for (int i=0; i<N; i++)
      sum[i]=a[i]+b[i];
}

__global__ void setup_kernel ( curandState * state, unsigned long seed )
{
   int id = threadIdx.x + blockIdx.x * 64;
   if (id<N){
      //seed, sequence, offset, state
      curand_init ( seed, id, 0, &state[id] );
   }
}

__global__ void generate( curandState* globalState, float * array1, float * array2, float * array3 )
{
   int ind = threadIdx.x;
   if (ind <N){
      curandState localState = globalState[ind];
      float RANDOM = curand_uniform( &localState );
      array1[ind] = RANDOM;
      array2[ind] = (RANDOM+5)/123;
      array3[ind] = RANDOM+2;
      globalState[ind] = localState;
   }
}

int main()
{
         curandState* devStates;
   float * a, * b, * sum, * vec;
   
   cudaMalloc ( &a, N*sizeof(float));
   cudaMalloc ( &b, N*sizeof(float));
   cudaMalloc ( &sum, N*sizeof(float));
   cudaMalloc ( &vec, N*sizeof(float));
   cudaMalloc ( &devStates, N*sizeof( curandState ));

   setup_kernel <<< N/256+1, 256 >>> (a, time(NULL));
   generate <<< N/256+1, 256 >>> ( devStates, a, b );

   #pragma acc declare device_resident (a[N], b[N], sum[N], vec[N])
   vecAdd (a,b,sum);

   cublasInit();
   cublasSaxpy(N, 2.0, sum, 2, vec, 1);
   cublasShutdown();
   
   #pragma acc host_data use_device (sum);
      
   for (int i=0; i<N; i++)
      printf(sum[i],"\n");
   return 0;
}


I compile it with:
Code:

pgcc -acc -I/opt/pgi/linux86-64/2012/cuda/4.2/include -Minfo=accel -L /opt/pgi/linux86-64/2012/cuda/4.2/lib64 -lcurand -ta=nvidia interop.c


And recieve the following error:

Code:

PGC-F-0249-#error --  --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- (/opt/pgi/linux86-64/2012/cuda/4.2/include/host_defines.h: 128)
PGC/x86-64 Linux 12.10-0: compilation aborted


Thanks a lot!
Back to top
View user's profile
mkcolg



Joined: 30 Jun 2004
Posts: 4996
Location: The Portland Group Inc.
PostPosted: Wed Dec 19, 2012 10:32 am    Post subject: Reply with quote
Hi i_alex2004,

Unfortunately, NVIDIA hasn't updated their header files to allow pgcc to compile them. Also, pgcc doesn't support CUDA C extensions. Our C++ compiler, pgcpp, does but only when targeting x86, not NVIDIA GPUs.

What this means is that you need to compile your CUDA C code with nvcc, and OpenACC code with PGI, and the two can't be mixed in the same file. The objects and device pointers are interoperable.

Hope this helps,
Mat
Back to top
View user's profile
Display posts from previous:   
Post new topic   Reply to topic    PGI User Forum Forum Index -> Accelerator Programming All times are GMT - 7 Hours
Page 1 of 1

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © 2001, 2002 phpBB Group